Search NASA⌕ Search

SEARCH · Search NASA

Results for “predictor”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 739 records · Page 41

Multi-institutional Nomogram Predicting Survival Free From Salvage Whole Brain Radiation After Radiosurgery in Patients With Brain Metastases

Purpose: Optimal patient selection for stereotactic radiosurgery (SRS) as the initial treatment for brain metastases is complicated and controversial. This study aimed to develop a nomogram that predicts survival without salvage whole brain radiation therapy (WBRT) after upfront SRS. Methods and Materials: Multi-institutional data were analyzed from 895 patients with 2095 lesions treated with SRS without prior or planned WBRT. Cox proportional hazards regression model was used to identify independent pre-SRS predictors of WBRT-free survival, which were integrated to build a nomogram that was subjected to bootstrap validation. Results: Median WBRT-free survival was 8 months (range, 0.1-139 months). Significant independent predictors for inferior WBRT-free survival were age (hazard ratio [HR] 1.1 for each 10-year increase), HER2(−) breast cancer (HR 1.6 relative to other histologic features), colorectal cancer (HR 1.4 relative to other histologic features), increasing number of brain metastases (HR 1.09, 1.32, 1.37, and 1.87 for 2, 3, 4, and 5+ lesions, respectively), presence of neurologic symptoms (HR 1.26), progressive systemic disease (HR 1.35), and increasing extracranial disease burden (HR 1.31 for oligometastatic and HR 1.56 for widespread). Additionally, HER2(+) breast cancer (HR 0.81) and melanoma (HR 1.11) trended toward significance. The independently weighted hazard ratios were used to create a nomogram to display estimated probabilities of 6-month and 12-month WBRT-free survival with a corrected Harrell's C concordance statistic of 0.62. Conclusions: Our nomogram can be used at initial evaluation to help select patients best suited for upfront SRS for brain metastases while reducing expense and morbidity in patients who derive minimal or no benefit.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

National Cancer Database Analysis of Proton Versus Photon Radiation Therapy in Non-Small Cell Lung Cancer

Purpose: To analyze outcomes and predictors associated with proton radiation therapy for non-small cell lung cancer (NSCLC) in the National Cancer Database. Methods and Materials: The National Cancer Database was queried to capture patients with stage I-IV NSCLC treated with thoracic radiation from 2004 to 2012. A logistic regression model was used to determine the predictors for utilization of proton radiation therapy. The univariate and multivariable association with overall survival were assessed by Cox proportional hazards models along with log–rank tests. A propensity score matching method was implemented to balance baseline covariates and eliminate selection bias. Results: A total of 243,822 patients (photon radiation therapy: 243,474; proton radiation therapy: 348) were included in the analysis. Patients in a ZIP code with a median income of <$46,000 per year were less likely to receive proton treatment, with the income cohort of $30,000 to $35,999 least likely to receive proton therapy (odds ratio 0.63 [95% confidence interval (CI) 0.44-0.90]; P=.011). On multivariate analysis of all patients, non-proton therapy was associated with significantly worse survival compared with proton therapy (hazard ratio 1.21 [95% CI 1.06-1.39]; P<.01). On propensity matched analysis, proton radiation therapy (n=309) was associated with better 5-year overall survival compared with non-proton radiation therapy (n=1549), 22% versus 16% (P=.025). For stage II and III patients, non-proton radiation therapy was associated with worse survival compared with proton radiation therapy (hazard ratio 1.35 [95% CI 1.10-1.64], P<.01). Conclusions: Thoracic radiation with protons is associated with better survival in this retrospective analysis; further validation in the randomized setting is needed to account for any imbalances in patient characteristics, including positron emission tomography–computed tomography staging.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Sentinel Lymph Node Dissection to Select Clinically Node-negative Prostate Cancer Patients for Pelvic Radiation Therapy: Effect on Biochemical Recurrence and Systemic Progression

Purpose: To assess the efficacy of robotic-assisted laparoscopic sentinel lymph node (SLN) dissection (SLND) to select those patients with prostate cancer (PCa) who would benefit from additional pelvic external beam radiation therapy and long-term androgen deprivation therapy (ADT). Methods and Materials: Radioisotope-guided SLND was performed in 224 clinically node-negative patients scheduled to undergo external beam radiation therapy. Patients with histologically positive SLNs (pN1) were also offered radiation therapy to the pelvic lymph nodes, combined with 3 years of ADT. Biochemical recurrence (BCR), overall survival, and metastasis-free (including pelvic and nonregional lymph nodes) survival (MFS) rates were retrospectively calculated. The Briganti and Kattan nomogram predictions were compared with the observed pN status and BCR. Results: The median prostate-specific antigen (PSA) value was 15.4 ng/mL (interquartile range [IQR] 8-29). A total number of 834 SLNs (median 3 per patient; IQR 2-5) were removed. Nodal metastases were diagnosed in 42% of the patients, with 150 SLNs affected (median 1; IQR 1-2). The 5-year BCR-free and MFS rates for pN0 patients were 67.9% and 87.8%, respectively. The corresponding values for pN1 patients were 43% and 66.6%. The PSA level and number of removed SLNs were independent predictors of BCR and MFS, and pN status was an additional independent predictor of BCR. The 5-year overall survival rate was 97.6% and correlated only with pN status. The predictive accuracy of the Briganti nomogram was 0.665. Patients in the higher quartiles of Kattan nomogram prediction of BCR had better than expected outcomes. The complication rate from SLND was 8.9%. Conclusions: For radioisotope-guided SLND, the high staging accuracy is accompanied by low morbidity. The better than expected outcomes observed in the lower quartiles of BCR prediction suggest a role for SLN biopsy as a potential selection tool for the addition of pelvic radiation therapy and ADT intensification in pN1 patients.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Radiation Dose–Dependent Hippocampal Atrophy Detected With Longitudinal Volumetric Magnetic Resonance Imaging

Purpose: After radiation therapy (RT) to the brain, patients often experience memory impairment, which may be partially mediated by damage to the hippocampus. Hippocampal sparing in RT planning is the subject of recent and ongoing clinical trials. Calculating appropriate hippocampal dose constraints would be improved by efficient in vivo measurements of hippocampal damage. In this study we sought to determine whether brain RT was associated with dose-dependent hippocampal atrophy. Methods and Materials: Hippocampal volume was measured with magnetic resonance imaging (MRI) in 52 patients who underwent fractionated, partial brain RT for primary brain tumors. Study patients had high-resolution, 3-dimensional volumetric MRI before and 1 year after RT. Images were processed using software with clearance from the US Food and Drug Administration and Conformité Européene marking for automated measurement of hippocampal volume. Automated results were inspected visually for accuracy. Tumor and surgical changes were censored. Mean hippocampal dose was tested for correlation with hippocampal atrophy 1 year after RT. Average hippocampal volume change was also calculated for hippocampi receiving high (>40 Gy) or low (<10 Gy) mean RT dose. A multivariate analysis was conducted with linear mixed-effects modeling to evaluate other potential predictors of hippocampal volume change, including patient (random effect), age, hemisphere, sex, seizure history, and baseline volume. Statistical significance was evaluated at α = 0.05. Results: Mean hippocampal dose was significantly correlated with hippocampal volume loss (r=−0.24, P=.03). Mean hippocampal volume was significantly reduced 1 year after high-dose RT (mean −6%, P=.009) but not after low-dose RT. In multivariate analysis, both RT dose and patient age were significant predictors of hippocampal atrophy (P<.01). Conclusions: The hippocampus demonstrates radiation dose–dependent atrophy after treatment for brain tumors. Quantitative MRI is a noninvasive imaging technique capable of measuring radiation effects on intracranial structures. This technique could be investigated as a potential biomarker for development of reliable dose constraints for improved cognitive outcomes.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Does Unintentional Splenic Radiation Predict Outcomes After Pancreatic Cancer Radiation Therapy?

Purpose: To determine whether severity of lymphopenia is dependent on radiation dose and fractional volume of spleen irradiated unintentionally during definitive chemoradiation (CRT) in patients with locally advanced pancreatic cancer (LAPC). Methods: 177 patients with LAPC received induction chemotherapy (mainly gemcitabine-based regimens) followed by CRT (median 50.4 Gy with concurrent capecitabine) from January 2006 to December 2012. Absolute lymphocyte count (ALC) was recorded at baseline, before CRT, and 2 to 10 weeks after CRT. Splenic dose-volume histogram (DVH) parameters were reported as mean splenic dose (MSD) and percentage of splenic volume receiving at least 5- (V5), 10- (V10), 15- (V15), and 20-Gy (V20) dose. Overall survival (OS) was analyzed with use of the Cox model, and development of post-CRT severe lymphopenia (ALC <0.5 K/UL) was assessed by multivariate logistic regression with use of baseline and treatment factors. Results: The median post-CRT ALC (0.68 K/UL; range, 0.13-2.72) was significantly lower than both baseline ALC (1.42 K/UL; range, 0.34-3.97; P<.0001) and pre-CRT ALC (1.32 K/UL, range 0.36-4.82; P<.0001). Post-CRT ALC <0.5 K/UL was associated with inferior OS on univariate analysis (median, 11.1 vs 15.3 months; P=.01) and multivariate analysis (hazard ratio = 1.66, P=.01). MSD (9.8 vs 6 Gy, P=.03), median V10 (32.6 vs 16%, P=.04), V15 (23.2 vs 9.5%, P=.03), and V20 (15.4 vs 4.6%, P=.02) were significantly higher in patients with severe lymphopenia than in those without. On multivariate analysis, postinduction lymphopenia (P<.001; odds ratio [OR] = 5.25) and MSD (P=.002; OR= 3.42) were independent predictors for the development of severe post-CRT lymphopenia. Conclusion: Severe post-CRT lymphopenia is an independent predictor of poor OS in LAPC patients receiving CRT. Higher splenic doses increase the risk for the development of severe post-CRT lymphopenia. When clinically indicated, assessment of splenic DVHs before the acceptance of treatment plans may minimize the risk of severe post-CRT lymphopenia.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Safety and Efficacy of Stereotactic Ablative Radiation Therapy for Renal Cell Carcinoma Extracranial Metastases

Purpose: Renal cell carcinoma is refractory to conventional radiation therapy but responds to higher doses per fraction. However, the dosimetric data and clinical factors affecting local control (LC) are largely unknown. We aimed to evaluate the safety and efficacy of stereotactic ablative radiation therapy (SAbR) for extracranial renal cell carcinoma metastases. Methods and Materials: We reviewed 175 metastatic lesions from 84 patients treated with SAbR between 2005 and 2015. LC and toxicity after SAbR were assessed with Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 and Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. Predictors of local failure were analyzed with χ{sup 2}, Kaplan-Meier, and log-rank tests. Results: In most cases (74%), SAbR was delivered with total doses of 40 to 60 Gy, 30 to 54 Gy, and 20 to 40 Gy in 5 fractions, 3 fractions, and a single fraction, respectively. The median biologically effective dose (BED) using the universal survival model was 134.5 Gy. The 1-year LC rate after SAbR was 91.2% (95% confidence interval, 84.9%-95.0%; median follow-up, 16.7 months). Local failures were associated with prior radiation therapy (hazard ratio [HR], 10.49; P<.0001), palliative-intent radiation therapy (HR, 4.63; P=.0189), spinal location (HR, 5.36; P=.0041), previous systemic therapy status (0-1 vs >1; HR, 3.52; P=.0217), and BED <115 Gy (HR, 3.45; P=.0254). Dose received by 99% of the target volume was the strongest dosimetric predictor for LC. Upon multivariate analysis, dose received by 99% of the target volume greater than BED of 98.7 Gy and systemic therapy status remained significant (HR, 0.12 and 3.64, with P=.0014 and P=.0472, respectively). Acute and late grade 3 toxicities attributed to SAbR were observed in 3 patients (1.7%) and 5 patients (2.9%), respectively. Conclusions: SAbR demonstrated excellent LC of metastatic renal cell carcinoma with a favorable safety profile when an adequate dose and coverage were applied. Multimodality treatment with surgery should be considered for reirradiation or vertebral metastasis. A higher radiation dose may be required in patients who received previous systemic therapies.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Reporting Quality of Randomized, Controlled Trials Evaluating Combined Chemoradiotherapy in Nasopharyngeal Carcinoma

Purpose: To comprehensively assess the reporting quality of randomized, controlled trials (RCTs) in nasopharyngeal carcinoma (NPC), and to identify significant predictors of quality. Methods and Materials: Two investigators searched MEDLINE and EMBASE for RCTs published between January 1988 and December 2015 that assessed the effect of combined chemoradiotherapy for NPC. The overall quality of each report was assessed using a 28-point overall quality score (OQS) based on the 2010 Consolidated Standards of Reporting Trials (CONSORT) statement. To provide baseline data for further evaluation, we also investigated the reporting quality of certain important issues in detail, including key methodologic items (allocation concealment, blinding, intention-to-treat principle), endpoints, follow-up, subgroup analyses, and adverse events. Results: We retrieved 24 relevant RCTs including 6591 patients. Median 2010 OQS was 15.5 (range, 10-24). Half of the items in the 2010 OQS were poorly reported in at least 40% of trials. Multivariable regression models revealed that publication after 2010 and high impact factor were significant predictors of improved 2010 OQS. Additionally, many issues that we consider significant were not reported adequately. Conclusions: Despite publication of the CONSORT statement more than a decade ago, overall reporting quality for RCTs in NPC was unsatisfactory. Additionally, substantial selectivity and heterogeneity exists in reporting of certain crucial issues. This survey provides the first prompt for NPC trial investigators to improve reporting quality according to the CONSORT statement; increased scrutiny and diligence by editors and peer reviewers is also required.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Histologic Appearance After Preoperative Radiation Therapy for Soft Tissue Sarcoma: Assessment of the European Organization for Research and Treatment of Cancer–Soft Tissue and Bone Sarcoma Group Response Score

Purpose: To critically assess the prognostic value of the European Organization for Research and Treatment of Cancer–Soft Tissue and Bone Sarcoma Group (EORTC-STBSG) response score and define histologic appearance after preoperative radiation therapy (RT) for soft tissue sarcoma (STS). Methods and Materials: For a cohort of 100 patients with STS of the extremity/trunk treated at our institution with preoperative RT followed by resection, 2 expert sarcoma pathologists evaluated the resected specimens for percent residual viable cells, necrosis, hyalinization/fibrosis, and infarction. The EORTC response score and other predictors of recurrence-free survival (RFS) and overall survival (OS) were assessed by Kaplan-Meier and proportional hazard models. Results: Median tumor size was 7.5 cm; 92% were intermediate or high grade. Most common histologies were unclassified sarcoma (34%) and myxofibrosarcoma (25%). Median follow-up was 60 months. The 5-year local recurrence rate was 5%, 5-year RFS was 68%, and 5-year OS was 75%. Distribution of cases according to EORTC response score tiers was as follows: no residual viable tumor for 9 cases (9% pathologic complete response); <1% viable tumor for 0, ≥1% to <10% for 9, ≥10% to <50% for 44, and ≥50% for 38. There was no association between EORTC-STBSG response score and RFS or OS. Conversely, hyalinization/fibrosis was a significant independent favorable predictor for RFS (hazard ratio 0.49, P=.007) and OS (hazard ratio 0.36, P=.02). Conclusion: Histologic evaluation after preoperative RT for STS showed a 9% pathologic complete response rate. The EORTC-STBSG response score and percent viable cells were not prognostic. Hyalinization/fibrosis was associated with favorable outcome, and if validated, may become a valid endpoint for neoadjuvant trials.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

In Vivo Brain Plaque and Tangle Burden Mediates the Association Between Diastolic Blood Pressure and Cognitive Functioning in Nondemented Adults

OBJECTIVE Growing evidence supports an association among increased blood pressure and: (a) poor cognitive performance in older adults, and (b) various biomarkers of increased Alzheimer's disease (AD) neuropathology. The objective of this study was to determine whether systolic blood pressure (SBP) and diastolic blood pressure (DBP) were significantly associated with cognitive functioning in non-demented adults, and to examine in vivo AD pathology as a possible mediator of this association. METHODS Positron emission tomography (PET) scans with 2-(1-{6-[(2-[F-18]fluoroethyl)(methyl)amino]-2-naphthyl}ethylidene)malononitrile (FDDNP) provide in vivo measurements of plaque and tangle burden. A total of 101 non-demented older subjects with blood pressure data and FDDNP-PET scans were drawn from a larger study of predictors of cognitive decline. A neuropsychological test battery was used to compute "global cognitive scores" (averaged across five key domains), which served as an index of general cognitive functioning. RESULTS Higher DBP (but not SBP) was significantly associated with lower cognitive scores, controlling for age, sex, antihypertensive medication use, and ApoE genotype (η2 = 0.06). However, this relationship was no longer significant after introducing FDDNP-PET binding as an additional covariate in the statistical models. In vivo plaque and tangle burden accounted for over 30% of the observed association between higher DBP and poorer cognitive performance. CONCLUSIONS By indicating a mediation of the relationship between DBP and cognitive functioning by FDDNP-PET binding, this study advances our understanding of some potential predictors of cognitive decline in non-demented adults, and underscores the importance of devising early multimodal interventions to more effectively combat degenerative brain disorders.

60 APPLIED LIFE SCIENCES↗

Conservative explicit local time-stepping schemes for the shallow water equations

Here we develop explicit local time-stepping (LTS) schemes with second and third order accuracy for the shallow water equations. The system is discretized in space by a C-grid staggering method, namely the TRiSK scheme adopted in MPAS-Ocean, a global ocean model with the capability of resolving multiple resolutions within a single simulation. The time integration is designed based on the strong stability preserving Runge–Kutta (SSP-RK) methods, but different time step sizes can be used in different regions of the domain through the coupling of coarse-fine time discretizations on the interface, and are only restricted by respective local CFL conditions. The proposed LTS schemes are of predictor–corrector type in which the predictors are constructed based on Taylor series expansions and SSP-RK stepping algorithms. The schemes preserve some important physical quantities in the discrete sense, such as exact conservation of the mass and potential vorticity and conservation of the total energy within time truncation errors. Moreover, they inherit the natural parallelism of the original explicit global time-stepping schemes. Extensive numerical tests are presented to illustrate the performance of the proposed algorithms.

58 GEOSCIENCES↗

Comparison of four learning-based methods for predicting groundwater redox status

Knowing the location where groundwater denitrification occurs, or by proxy the groundwater redox status (oxic, mixed, and anoxic), is valuable information for assessing and managing potential agricultural land-use impacts on freshwater quality. We compare the efficacy of supervised (Linear Discriminant Analysis LDA; Boosted Regression Trees, BRT; and Random Forest, RF) and unsupervised (Modified Self-Organizing Map, MSOM) learning-based methods to predict groundwater redox status in the agriculturally dominated Tasman, Waikato, and Wellington regions of New Zealand. Thresholds applied to regional groundwater-quality samples provide redox status variables and learn heuristics constrained by these variables and applied to spatial factors (climate, elevation, geologic, hydrology soils, and well depth) identify optimal sets of regional predictor variables. A split-sample approach is used to train and test the learning methods ability to predict redox status using the optimal predictor variables. Overall, the supervised methods demonstrate a prediction bias toward oxic conditions and inability to perform statistically well when using independent regional data; for example, consider kappa statistics for BRT (Tasman: 0.42, Waikato: 0.38, Wellington: 0.17), RF (Tasman: 0.42, Waikato: 0.47, Wellington: 0.17 and LDA (Tasman: 0.46, Waikato: 0.32, Wellington: 0.17). By contrast, the unsupervised method performs statistically well when predicting oxic, mixed, and anoxic conditions and corresponding depths when using independent regional data; for example, consider MSOM kappa statistics for Tasman: 0.78, Waikato: 0.80, Wellington: 0.76. The unsupervised learning method provides the added benefits of being (1) able to combine predictions into 3D regional anoxic probability plots for interpreting the spatial influence of paleosols and groundwater flowpaths on redox status, and (2) readily extended to map 3D redox status across New Zealand and other countries despite data bias and sparsity.

machine learning algorithms, Random Forest, booste↗

Lithium-ion battery physics and statistics-based state of health model

A pseudo-2d model using COMSOL Multiphysics® software is developed to simulate performance and performance degradation of Li-ion batteries consisting of layered and olivine cathodes with graphite anode when subjected to peak shaving grid service. Multiple degradation pathways are considered, including solid electrolyte interphase (SEI) formation and breakdown at the anode, cathode dissolution and its synergistic effect on SEI formation at the anode. The model is validated by simulating commercial cylindrical cell performance. A global model is developed to simulate performance across all chemistries, along with individual chemistry models using global model parameters as initial values. There is good agreement between these models for various optimization parameters such as SEI equilibrium potential, cathode dissolution exchange current density, solvent diffusivity in the SEI and SEI ionic conductivity. To circumvent time constraints related to the COMSOL model, a 0d global model is developed which fits data well and provides more clarity on differences in cathode dissolution exchange current density. Again, good agreement for various optimization parameters is obtained among the COMSOL global & individual chemistry models and the 0-d model. The lessons learned from the physics-based model is used to develop a top down statistics-based model using current, voltage and anode volumetric change per mole lithium intercalated, along with their interactions as degradation predictors. This model predicts out of sample degradation for multiple grid services and electric vehicle drive cycle with high accuracy and provides the pathway to develop an efficient battery management system combining machine learning and findings from physics-based computationally intensive algorithms.

Crawford, Aladsair J.↗

Bilateral implant reconstruction does not affect the quality of postmastectomy radiation therapy

To determine if the presence of bilateral implants, in addition to other anatomic and treatment-related variables, affects coverage of the target volume and dose to the heart and lung in patients receiving postmastectomy radiation therapy (PMRT). A total of 197 consecutive women with breast cancer underwent mastectomy and immediate tissue expander (TE) placement, with or without exchange for a permanent implant (PI) before radiation therapy at our center. PMRT was delivered with 2 tangential beams + supraclavicular lymph node field (50 Gy). Patients were grouped by implant number: 51% unilateral (100) and 49% bilateral (97). The planning target volume (PTV) (defined as implant + chest wall + nodes), heart, and ipsilateral lung were contoured and the following parameters were abstracted from dose-volume histogram (DVH) data: PTV D{sub 95%} > 98%, Lung V{sub 20}Gy > 30%, and Heart V{sub 25}Gy > 5%. Univariate (UVA) and multivariate analyses (MVA) were performed to determine the association of variables with these parameters. The 2 groups were well balanced for implant type and volume, internal mammary node (IMN) treatment, and laterality. In the entire cohort, 90% had PTV D{sub 95%} > 98%, indicating excellent coverage of the chest wall. Of the patients, 27% had high lung doses (V{sub 20}Gy > 30%) and 16% had high heart doses (V{sub 25}Gy > 5%). No significant factors were associated with suboptimal PTV coverage. On MVA, IMN treatment was found to be highly associated with high lung and heart doses (both p < 0.0001), but implant number was not (p = 0.54). In patients with bilateral implants, IMN treatment was the only predictor of dose to the contralateral implant (p = 0.001). In conclusion, bilateral implants do not compromise coverage of the target volume or increase lung and heart dose in patients receiving PMRT. The most important predictor of high lung and heart doses in patients with implant-based reconstruction, whether unilateral or bilateral, is treatment of the IMNs. Refinement of radiation techniques in reconstructed patients who require comprehensive nodal irradiation is warranted.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Volumetric tumor burden and its effect on brachial plexus dosimetry in head and neck intensity-modulated radiotherapy

To determine the effect of gross tumor volume of the primary (GTV-P) and nodal (GTV-N) disease on planned radiation dose to the brachial plexus (BP) in head and neck intensity-modulated radiotherapy (IMRT). Overall, 75 patients underwent definitive IMRT to a median total dose of 69.96 Gy in 33 fractions. The right BP and left BP were prospectively contoured as separate organs at risk. The GTV was related to BP dose using the unpaired t-test. Receiver operating characteristics curves were constructed to determine optimized volumetric thresholds of GTV-P and GTV-N corresponding to a maximum BP dose cutoff of > 66 Gy. Multivariate analyses were performed to account for factors associated with a higher maximal BP dose. A higher maximum BP dose (> 66 vs ≤ 66 Gy) correlated with a greater mean GTV-P (79.5 vs 30.8 cc; p = 0.001) and ipsilateral GTV-N (60.6 vs 19.8 cc; p = 0.014). When dichotomized by the optimized nodal volume, patients with an ipsilateral GTV-N ≥ 4.9 vs < 4.9 cc had a significant difference in maximum BP dose (64.2 vs 59.4 Gy; p = 0.001). Multivariate analysis confirmed that an ipsilateral GTV-N ≥ 4.9 cc was an independent predictor for the BP to receive a maximal dose of > 66 Gy when adjusted individually for BP volume, GTV-P, the use of a low anterior neck field technique, total planned radiation dose, and tumor category. Although both the primary and the nodal tumor volumes affected the BP maximal dose, the ipsilateral nodal tumor volume (GTV-N ≥ 4.9 cc) was an independent predictor for high maximal BP dose constraints in head and neck IMRT.

61 RADIATION PROTECTION AND DOSIMETRY↗

Radiation-induced hypothyroidism in patients with breast cancer: a retrospective analysis of 243 cases

This study aims to estimate the incidence of hypothyroidism (HT) and to evaluate the predictors affecting the development of HT after radiotherapy (RT) for breast cancer, with a focus on radiation dose-volumetric parameters. Between 2009 and 2015, 243 patients undergoing RT for breast cancer were retrospectively analyzed. Free triiodothyronine (FT3), free thyroxin (FT4), and thyrotropin (TSH) were monitored before and after RT. The relation between the doses to thyroid gland (D{sub mean}, D{sub max}, D{sub min}), percentage of thyroid volume receiving > 10 Gy, 20 Gy, 30 Gy, 40 Gy, and 50 Gy (V10 to V50), absolute thyroid volume, and HT were analyzed. The risk of HT according to radiation fields and the other clinic factors were also evaluated. The median follow-up was 41 (range; 6 to 130) months. Sixty-seven percent of the patients received RT to the breast/chest wall and ipsilateral supraclavicular fossa. Of 243 patients, 51 (21%) were diagnosed with HT. The median time to the onset of HT was 27 (range; 5 to 64) months. There were no significant relationships observed between D{sub min} or V10 to V50 and HT. The surgery type, clinical stage, nodal status, RT field, D{sub mean}, and D{sub max} were statistically significant predictors for HT in univariate analysis. The D{sub mean} was the only prognostic factor predicting HT in multivariate analysis, and D{sub mean} > 21 Gy was a threshold value for the evolvement of HT. In this study, we present evidence that postoperative irradiation of patients with breast cancer may frequently lead to HT. Patients who have received RT for breast cancer, especially irradiation on the supraclavicular region, may require thyroid function screening after RT.

61 RADIATION PROTECTION AND DOSIMETRY↗

Explaining drivers of housing prices with nonlinear hedonic regressions

Housing markets play a critical role in shaping the spatial and demographic evolution of urban areas. Simulating housing price dynamics can enhance projections of future urban development outcomes. However, traditional hedonic regressions for housing prices, which neglect nonlinear interactions among explanatory variables, often exhibit limited predictive performance. While machine learning (ML) methods can provide a more flexible representation of the relationships between predictors, they are often regarded as “black boxes” due to their complexity and lack of transparency. Interpretable ML techniques provide a promising route by combining the flexibility of ML methods with approaches to analyze the relationships between inputs and outputs. In this study, we employ interpretable ML to analyze the patterns driving the housing market in Baltimore, Maryland, USA. We train an Artificial Neural Network (ANN) to predict Baltimore housing prices based on structural characteristics (e.g., home size, number of stories) and locational attributes (e.g., distance to the city center). We then conduct sensitivity and Partial Dependence Plot (PDP) analyses to interpret the fitted ANN model. We find that the ML model achieves higher predictive accuracy and explains 16 % more of housing price variance than a traditional linear regression model. The interpretable ML model also reveals more nuanced and realistic nonlinear relationships between housing sales price and predictors as well as interactive effects underlying Baltimore home price dynamics. For instance, while the linear model indicates a steady housing price increase over time, our interpretable ML model detects a post-2008 decline, with smaller properties experiencing the sharpest drop.

97 MATHEMATICS AND COMPUTING↗

Predicting industrial building energy consumption with statistical and machine-learning models informed by physical system parameters

The industrial sector consumes about one-third of global energy, making them a frequent target for energy use reduction. Variation in energy usage is observed with weather conditions, as space conditioning needs to change seasonally, and with production, energy-using equipment is directly tied to production rate. Previous models were based on engineering analyses of equipment and relied on site-specific details. Others consisted of single-variable regressors that did not capture all contributions to energy consumption. Further, new modeling techniques could be applied to rectify these weaknesses. Applying data from 45 different manufacturing plants obtained from industrial energy audits, a supervised machine-learning model is developed to create a general predictor for industrial building energy consumption. The model uses features of air enthalpy, solar radiation, and wind speed to predict weather-dependency; motor, steam, and compressed air system parameters to capture support equipment contributions; and operating schedule, production rate, number of employees, and floor area to determine production-dependency. Results showed that a model that used a linear regressor over a transformed feature space could outperform a support vector machine and utilize features more representative of physical systems. Using informed parameters to build a reliable predictor will more accurately characterize a manufacturing facility's energy savings opportunities.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗